In [1]:
import os
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Conv3D, MaxPooling3D, UpSampling3D, concatenate, Conv3DTranspose, BatchNormalization, Dropout, Lambda
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.layers import Activation, MaxPool2D, Concatenate

import scipy
from skimage import transform
from skimage import io

import numpy as np
from matplotlib import pyplot as plt
from tensorflow.keras import backend as K
from tensorflow.keras.utils import to_categorical
from sklearn.model_selection import train_test_split

import tifffile
!pip install patchify

Defaulting to user installation because normal site-packages is not writeable
Requirement already satisfied: patchify in ./.local/lib/python3.8/site-packages (0.2.3)
Requirement already satisfied: numpy<2,>=1 in /data/software/python-libs/3.0.0/lib/python3.8/site-packages (from patchify) (1.22.3)
In [2]:
def plot_all(image):
    index = 0
    count = 1
    fig = plt.figure(figsize = (15, 30))
    for x in range(1, 17):
        for y in range(1, 9):
            plt.subplot(16, 8, count).axis("off")
            plt.title("Count: " + str(count-1))
            plt.imshow(image[:,:,index])
            count += 1
            index += 1
In [3]:
''' Read Images/Masks '''
images = tifffile.imread('ivc_filter_images_84.tif')
masks = tifffile.imread("ivc_filter_masks_84.tif")
print(images.shape)
print(masks.shape)

(84, 256, 256, 128)
(84, 256, 256, 128)
In [4]:
''' Splits into training/testing '''
x_train_orig, x_test_orig, y_train_orig, y_test_orig = train_test_split(images, masks, test_size=0.20, random_state=7)
print(x_train_orig.shape)
print(y_train_orig.shape)
print(x_test_orig.shape)
print(y_test_orig.shape)

(67, 256, 256, 128)
(67, 256, 256, 128)
(17, 256, 256, 128)
(17, 256, 256, 128)
In [5]:
''' Creating Patches '''
import patchify
from skimage.util.shape import view_as_blocks, view_as_windows
from patchify import patchify, unpatchify

def to_patches(images, masks):
    image_patches = []
    mask_patches = []
    for img in images:
        img = img/255
        image_patches.append(view_as_blocks(img, (64, 64, 32)))
    for msk in masks:
        # Normalize to 0 or 1
        if msk.max() != 0:
            msk = msk / msk.max()
        mask_patches.append(view_as_blocks(msk, (64, 64, 32)))
    image_patches = np.reshape(image_patches, (-1, 64, 64, 32))
    mask_patches = np.reshape(mask_patches, (-1, 64, 64, 32))
    return image_patches, mask_patches
        
def patches_to_img(patches):
    temp = np.reshape(patches, (4, 4, 4, 64, 64, 32))
    end = unpatchify(temp, (256, 256, 128))
    return end
In [6]:
''' Split into Patches '''
x1_patches, y1_patches = to_patches(x_train_orig, y_train_orig)
x2_patches, y2_patches = to_patches(x_test_orig, y_test_orig)
print(np.asarray(x1_patches).shape)
print(np.asarray(y1_patches).shape)

(4288, 64, 64, 32)
(4288, 64, 64, 32)
In [7]:
''' Separate the IVCF Patches from non-IVCF Patches'''

x1_zeros = [] # holds the background patches in x_train
x1_ones = [] # holds the ivcf patches in x_train

y1_zeros = [] # holds the background patches in y_train
y1_ones = [] # holds the ivcf patches in y_train

x2_zeros = [] # Same as above but for testing set
x2_ones = []

y2_zeros = []
y2_ones = []

for index in range(np.asarray(x1_patches).shape[0]):
    if y1_patches[index].max() == 1:
        x1_ones.append(x1_patches[index])
        y1_ones.append(y1_patches[index])
    else:
        x1_zeros.append(x1_patches[index])
        y1_zeros.append(y1_patches[index])
        
for index in range(np.asarray(x2_patches).shape[0]):
    if y2_patches[index].max() == 1:
        x2_ones.append(x2_patches[index])
        y2_ones.append(y2_patches[index])
    else:
        x2_zeros.append(x2_patches[index])
        y2_zeros.append(y2_patches[index])
        
print(np.asarray(x1_zeros).shape)
print(np.asarray(x1_ones).shape)
print(np.asarray(y1_zeros).shape)
print(np.asarray(y1_ones).shape)

(4113, 64, 64, 32)
(175, 64, 64, 32)
(4113, 64, 64, 32)
(175, 64, 64, 32)
In [8]:
print(np.asarray(x2_zeros).shape)
print(np.asarray(x2_ones).shape)
print(np.asarray(y2_zeros).shape)
print(np.asarray(y2_ones).shape)

(1035, 64, 64, 32)
(53, 64, 64, 32)
(1035, 64, 64, 32)
(53, 64, 64, 32)
In [9]:
''' Create Training/Testing Set with 1:1 Ratio'''
max_len_train = np.asarray(x1_ones).shape[0]
max_len_test = np.asarray(x2_ones).shape[0]

x_train = x1_ones
y_train = y1_ones

# Thresh value for modulus
thresh1 = int(np.asarray(x1_zeros).shape[0] / np.asarray(x1_ones).shape[0])
thresh2 = int(np.asarray(x2_zeros).shape[0] / np.asarray(x2_ones).shape[0])
print(thresh1, thresh2)

for i, val in enumerate(x1_zeros):
    if i % thresh1 == 0 and np.asarray(x_train).shape[0] < max_len_train*2:
        x_train.append(x1_zeros[i])
        y_train.append(y1_zeros[i])


x_test = x2_ones
y_test = y2_ones

for i, val in enumerate(x2_zeros):
    if i % thresh2 == 0 and np.asarray(x_test).shape[0] < max_len_test*2:
        x_test.append(x2_zeros[i])
        y_test.append(y2_zeros[i])

23 19
In [10]:
''' Expand_dims and One Hot Encoding'''
x_train = np.expand_dims(np.asarray(x_train), axis = -1)
x_test = np.expand_dims(np.asarray(x_test), axis = -1)
y_train = to_categorical(np.asarray(y_train))
y_test = to_categorical(np.asarray(y_test))

print(np.asarray(x_train).shape)
print(np.asarray(y_train).shape)
print(np.asarray(x_test).shape)
print(np.asarray(y_test).shape)

(350, 64, 64, 32, 1)
(350, 64, 64, 32, 2)
(106, 64, 64, 32, 1)
(106, 64, 64, 32, 2)
In [11]:
#Define parameters for our model.
channels=1

LR = 0.0001
optim = keras.optimizers.Adam(LR)
In [12]:
def conv_block(input, num_filters):
    x = Conv3D(num_filters, 3, padding="same")(input)
    x = BatchNormalization()(x)   #Not in the original network. 
    x = Activation("relu")(x)

    x = Conv3D(num_filters, 3, padding="same")(x)
    x = BatchNormalization()(x)  #Not in the original network
    x = Activation("relu")(x)

    return x

#Encoder block: Conv block followed by maxpooling
def encoder_block(input, num_filters):
    x = conv_block(input, num_filters)
    p = MaxPooling3D((2, 2, 2))(x)
    return x, p   

#Decoder block
#skip features gets input from encoder for concatenation
def decoder_block(input, skip_features, num_filters):
    x = Conv3DTranspose(num_filters, (2, 2, 2), strides=2, padding="same")(input)
    x = Concatenate()([x, skip_features])
    x = conv_block(x, num_filters)
    return x

#Build Unet using the blocks
def build_unet(input_shape, n_classes):
    inputs = Input(input_shape)
    
    s1, p1 = encoder_block(inputs, 32)
    s2, p2 = encoder_block(p1, 64)
    # s3, p3 = encoder_block(p2, 128)
    # s4, p4 = encoder_block(p3, 256)

    b1 = conv_block(p2, 128) #Bridge

    # d1 = decoder_block(b1, s4, 256)
    # d2 = decoder_block(b1, s3, 128)
    d3 = decoder_block(b1, s2, 64)
    d4 = decoder_block(d3, s1, 32)

    if n_classes == 1:  #Binary
        activation = 'sigmoid'
    else:
        activation = 'softmax'

    outputs = Conv3D(n_classes, 1, padding="same", activation=activation)(d4)  #Change the activation based on n_classes
    print(activation)

    model = Model(inputs, outputs, name="U-Net")
    return model
In [13]:
METRICS = [
      tf.keras.metrics.TruePositives(name='tp'),
      tf.keras.metrics.FalsePositives(name='fp'),
      tf.keras.metrics.TrueNegatives(name='tn'),
      tf.keras.metrics.FalseNegatives(name='fn'), 
      tf.keras.metrics.BinaryAccuracy(name='accuracy'),
      tf.keras.metrics.Precision(name='precision'),
      tf.keras.metrics.Recall(name='recall'),
      tf.keras.metrics.AUC(name='auc'),
      tf.keras.metrics.AUC(name='prc', curve='PR'), # precision-recall curve
]
In [14]:
model = build_unet((64, 64, 32, 1), n_classes=2)
model.compile(optimizer = optim, loss=tf.keras.losses.CategoricalCrossentropy(), metrics=METRICS)
print(model.summary())

softmax
Model: "U-Net"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_1 (InputLayer)            [(None, 64, 64, 32,  0                                            
__________________________________________________________________________________________________
conv3d (Conv3D)                 (None, 64, 64, 32, 3 896         input_1[0][0]                    
__________________________________________________________________________________________________
batch_normalization (BatchNorma (None, 64, 64, 32, 3 128         conv3d[0][0]                     
__________________________________________________________________________________________________
activation (Activation)         (None, 64, 64, 32, 3 0           batch_normalization[0][0]        
__________________________________________________________________________________________________
conv3d_1 (Conv3D)               (None, 64, 64, 32, 3 27680       activation[0][0]                 
__________________________________________________________________________________________________
batch_normalization_1 (BatchNor (None, 64, 64, 32, 3 128         conv3d_1[0][0]                   
__________________________________________________________________________________________________
activation_1 (Activation)       (None, 64, 64, 32, 3 0           batch_normalization_1[0][0]      
__________________________________________________________________________________________________
max_pooling3d (MaxPooling3D)    (None, 32, 32, 16, 3 0           activation_1[0][0]               
__________________________________________________________________________________________________
conv3d_2 (Conv3D)               (None, 32, 32, 16, 6 55360       max_pooling3d[0][0]              
__________________________________________________________________________________________________
batch_normalization_2 (BatchNor (None, 32, 32, 16, 6 256         conv3d_2[0][0]                   
__________________________________________________________________________________________________
activation_2 (Activation)       (None, 32, 32, 16, 6 0           batch_normalization_2[0][0]      
__________________________________________________________________________________________________
conv3d_3 (Conv3D)               (None, 32, 32, 16, 6 110656      activation_2[0][0]               
__________________________________________________________________________________________________
batch_normalization_3 (BatchNor (None, 32, 32, 16, 6 256         conv3d_3[0][0]                   
__________________________________________________________________________________________________
activation_3 (Activation)       (None, 32, 32, 16, 6 0           batch_normalization_3[0][0]      
__________________________________________________________________________________________________
max_pooling3d_1 (MaxPooling3D)  (None, 16, 16, 8, 64 0           activation_3[0][0]               
__________________________________________________________________________________________________
conv3d_4 (Conv3D)               (None, 16, 16, 8, 12 221312      max_pooling3d_1[0][0]            
__________________________________________________________________________________________________
batch_normalization_4 (BatchNor (None, 16, 16, 8, 12 512         conv3d_4[0][0]                   
__________________________________________________________________________________________________
activation_4 (Activation)       (None, 16, 16, 8, 12 0           batch_normalization_4[0][0]      
__________________________________________________________________________________________________
conv3d_5 (Conv3D)               (None, 16, 16, 8, 12 442496      activation_4[0][0]               
__________________________________________________________________________________________________
batch_normalization_5 (BatchNor (None, 16, 16, 8, 12 512         conv3d_5[0][0]                   
__________________________________________________________________________________________________
activation_5 (Activation)       (None, 16, 16, 8, 12 0           batch_normalization_5[0][0]      
__________________________________________________________________________________________________
conv3d_transpose (Conv3DTranspo (None, 32, 32, 16, 6 65600       activation_5[0][0]               
__________________________________________________________________________________________________
concatenate (Concatenate)       (None, 32, 32, 16, 1 0           conv3d_transpose[0][0]           
                                                                 activation_3[0][0]               
__________________________________________________________________________________________________
conv3d_6 (Conv3D)               (None, 32, 32, 16, 6 221248      concatenate[0][0]                
__________________________________________________________________________________________________
batch_normalization_6 (BatchNor (None, 32, 32, 16, 6 256         conv3d_6[0][0]                   
__________________________________________________________________________________________________
activation_6 (Activation)       (None, 32, 32, 16, 6 0           batch_normalization_6[0][0]      
__________________________________________________________________________________________________
conv3d_7 (Conv3D)               (None, 32, 32, 16, 6 110656      activation_6[0][0]               
__________________________________________________________________________________________________
batch_normalization_7 (BatchNor (None, 32, 32, 16, 6 256         conv3d_7[0][0]                   
__________________________________________________________________________________________________
activation_7 (Activation)       (None, 32, 32, 16, 6 0           batch_normalization_7[0][0]      
__________________________________________________________________________________________________
conv3d_transpose_1 (Conv3DTrans (None, 64, 64, 32, 3 16416       activation_7[0][0]               
__________________________________________________________________________________________________
concatenate_1 (Concatenate)     (None, 64, 64, 32, 6 0           conv3d_transpose_1[0][0]         
                                                                 activation_1[0][0]               
__________________________________________________________________________________________________
conv3d_8 (Conv3D)               (None, 64, 64, 32, 3 55328       concatenate_1[0][0]              
__________________________________________________________________________________________________
batch_normalization_8 (BatchNor (None, 64, 64, 32, 3 128         conv3d_8[0][0]                   
__________________________________________________________________________________________________
activation_8 (Activation)       (None, 64, 64, 32, 3 0           batch_normalization_8[0][0]      
__________________________________________________________________________________________________
conv3d_9 (Conv3D)               (None, 64, 64, 32, 3 27680       activation_8[0][0]               
__________________________________________________________________________________________________
batch_normalization_9 (BatchNor (None, 64, 64, 32, 3 128         conv3d_9[0][0]                   
__________________________________________________________________________________________________
activation_9 (Activation)       (None, 64, 64, 32, 3 0           batch_normalization_9[0][0]      
__________________________________________________________________________________________________
conv3d_10 (Conv3D)              (None, 64, 64, 32, 2 66          activation_9[0][0]               
==================================================================================================
Total params: 1,357,954
Trainable params: 1,356,674
Non-trainable params: 1,280
__________________________________________________________________________________________________
None
In [15]:
'''Checks'''
print("Input shape", model.input_shape)
print("Output shape", model.output_shape)
print("-------------------")

Input shape (None, 64, 64, 32, 1)
Output shape (None, 64, 64, 32, 2)
-------------------
In [16]:
history=model.fit(x_train, y_train,
        validation_data=(x_test, y_test),
        batch_size=1,
        epochs=100,
        shuffle=True,
        verbose=1)     

Epoch 1/100
350/350 [==============================] - 24s 57ms/step - loss: 0.1496 - tp: 22760432.8917 - fp: 307872.3647 - tn: 22760432.8917 - fn: 307872.3647 - accuracy: 0.9826 - precision: 0.9826 - recall: 0.9826 - auc: 0.9920 - prc: 0.9885 - val_loss: 0.0551 - val_tp: 13754917.0000 - val_fp: 138715.0000 - val_tn: 13754917.0000 - val_fn: 138715.0000 - val_accuracy: 0.9900 - val_precision: 0.9900 - val_recall: 0.9900 - val_auc: 0.9936 - val_prc: 0.9925
Epoch 2/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0481 - tp: 22901040.7635 - fp: 167255.4729 - tn: 22901040.7635 - fn: 167255.4729 - accuracy: 0.9927 - precision: 0.9927 - recall: 0.9927 - auc: 0.9987 - prc: 0.9984 - val_loss: 0.0970 - val_tp: 13308297.0000 - val_fp: 585335.0000 - val_tn: 13308297.0000 - val_fn: 585335.0000 - val_accuracy: 0.9579 - val_precision: 0.9579 - val_recall: 0.9579 - val_auc: 0.9948 - val_prc: 0.9939
Epoch 3/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0372 - tp: 22886107.6581 - fp: 182192.9715 - tn: 22886107.6581 - fn: 182192.9715 - accuracy: 0.9915 - precision: 0.9915 - recall: 0.9915 - auc: 0.9990 - prc: 0.9987 - val_loss: 0.0872 - val_tp: 13310413.0000 - val_fp: 583219.0000 - val_tn: 13310413.0000 - val_fn: 583219.0000 - val_accuracy: 0.9580 - val_precision: 0.9580 - val_recall: 0.9580 - val_auc: 0.9958 - val_prc: 0.9954
Epoch 4/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0261 - tp: 22937332.9202 - fp: 130967.6439 - tn: 22937332.9202 - fn: 130967.6439 - accuracy: 0.9942 - precision: 0.9942 - recall: 0.9942 - auc: 0.9994 - prc: 0.9993 - val_loss: 0.0255 - val_tp: 13809303.0000 - val_fp: 84329.0000 - val_tn: 13809303.0000 - val_fn: 84329.0000 - val_accuracy: 0.9939 - val_precision: 0.9939 - val_recall: 0.9939 - val_auc: 0.9988 - val_prc: 0.9985
Epoch 5/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0232 - tp: 22939500.3761 - fp: 128801.6581 - tn: 22939500.3761 - fn: 128801.6581 - accuracy: 0.9938 - precision: 0.9938 - recall: 0.9938 - auc: 0.9995 - prc: 0.9995 - val_loss: 0.1819 - val_tp: 12530750.0000 - val_fp: 1362882.0000 - val_tn: 12530750.0000 - val_fn: 1362882.0000 - val_accuracy: 0.9019 - val_precision: 0.9019 - val_recall: 0.9019 - val_auc: 0.9796 - val_prc: 0.9801
Epoch 6/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0200 - tp: 22938080.2764 - fp: 130216.5071 - tn: 22938080.2764 - fn: 130216.5071 - accuracy: 0.9943 - precision: 0.9943 - recall: 0.9943 - auc: 0.9996 - prc: 0.9996 - val_loss: 0.0185 - val_tp: 13818924.0000 - val_fp: 74708.0000 - val_tn: 13818924.0000 - val_fn: 74708.0000 - val_accuracy: 0.9946 - val_precision: 0.9946 - val_recall: 0.9946 - val_auc: 0.9993 - val_prc: 0.9991
Epoch 7/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0178 - tp: 22954955.1681 - fp: 113344.7521 - tn: 22954955.1681 - fn: 113344.7521 - accuracy: 0.9950 - precision: 0.9950 - recall: 0.9950 - auc: 0.9997 - prc: 0.9996 - val_loss: 0.0173 - val_tp: 13825614.0000 - val_fp: 68018.0000 - val_tn: 13825614.0000 - val_fn: 68018.0000 - val_accuracy: 0.9951 - val_precision: 0.9951 - val_recall: 0.9951 - val_auc: 0.9994 - val_prc: 0.9993
Epoch 8/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0170 - tp: 22943573.6980 - fp: 124726.6154 - tn: 22943573.6980 - fn: 124726.6154 - accuracy: 0.9945 - precision: 0.9945 - recall: 0.9945 - auc: 0.9997 - prc: 0.9997 - val_loss: 0.0235 - val_tp: 13804553.0000 - val_fp: 89079.0000 - val_tn: 13804553.0000 - val_fn: 89079.0000 - val_accuracy: 0.9936 - val_precision: 0.9936 - val_recall: 0.9936 - val_auc: 0.9987 - val_prc: 0.9984
Epoch 9/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0130 - tp: 22975510.2450 - fp: 92791.7236 - tn: 22975510.2450 - fn: 92791.7236 - accuracy: 0.9959 - precision: 0.9959 - recall: 0.9959 - auc: 0.9998 - prc: 0.9998 - val_loss: 0.0320 - val_tp: 13793613.0000 - val_fp: 100019.0000 - val_tn: 13793613.0000 - val_fn: 100019.0000 - val_accuracy: 0.9928 - val_precision: 0.9928 - val_recall: 0.9928 - val_auc: 0.9974 - val_prc: 0.9966
Epoch 10/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0119 - tp: 22979746.6097 - fp: 88551.7436 - tn: 22979746.6097 - fn: 88551.7436 - accuracy: 0.9962 - precision: 0.9962 - recall: 0.9962 - auc: 0.9998 - prc: 0.9998 - val_loss: 0.0168 - val_tp: 13824360.0000 - val_fp: 69272.0000 - val_tn: 13824360.0000 - val_fn: 69272.0000 - val_accuracy: 0.9950 - val_precision: 0.9950 - val_recall: 0.9950 - val_auc: 0.9992 - val_prc: 0.9990
Epoch 11/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0115 - tp: 22982111.1937 - fp: 86187.7664 - tn: 22982111.1937 - fn: 86187.7664 - accuracy: 0.9960 - precision: 0.9960 - recall: 0.9960 - auc: 0.9999 - prc: 0.9998 - val_loss: 0.0162 - val_tp: 13827316.0000 - val_fp: 66316.0000 - val_tn: 13827316.0000 - val_fn: 66316.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9992 - val_prc: 0.9989
Epoch 12/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0113 - tp: 22978830.1225 - fp: 89472.4530 - tn: 22978830.1225 - fn: 89472.4530 - accuracy: 0.9960 - precision: 0.9960 - recall: 0.9960 - auc: 0.9999 - prc: 0.9998 - val_loss: 0.0211 - val_tp: 13817284.0000 - val_fp: 76348.0000 - val_tn: 13817284.0000 - val_fn: 76348.0000 - val_accuracy: 0.9945 - val_precision: 0.9945 - val_recall: 0.9945 - val_auc: 0.9986 - val_prc: 0.9982
Epoch 13/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0093 - tp: 22988711.7236 - fp: 79586.4558 - tn: 22988711.7236 - fn: 79586.4558 - accuracy: 0.9967 - precision: 0.9967 - recall: 0.9967 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0162 - val_tp: 13826227.0000 - val_fp: 67405.0000 - val_tn: 13826227.0000 - val_fn: 67405.0000 - val_accuracy: 0.9951 - val_precision: 0.9951 - val_recall: 0.9951 - val_auc: 0.9992 - val_prc: 0.9989
Epoch 14/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0091 - tp: 22993034.9573 - fp: 75263.9202 - tn: 22993034.9573 - fn: 75263.9202 - accuracy: 0.9968 - precision: 0.9968 - recall: 0.9968 - auc: 0.9998 - prc: 0.9998 - val_loss: 0.0222 - val_tp: 13811894.0000 - val_fp: 81738.0000 - val_tn: 13811894.0000 - val_fn: 81738.0000 - val_accuracy: 0.9941 - val_precision: 0.9941 - val_recall: 0.9941 - val_auc: 0.9984 - val_prc: 0.9980
Epoch 15/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0083 - tp: 22997843.8063 - fp: 70457.2137 - tn: 22997843.8063 - fn: 70457.2137 - accuracy: 0.9969 - precision: 0.9969 - recall: 0.9969 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0190 - val_tp: 13823618.0000 - val_fp: 70014.0000 - val_tn: 13823618.0000 - val_fn: 70014.0000 - val_accuracy: 0.9950 - val_precision: 0.9950 - val_recall: 0.9950 - val_auc: 0.9986 - val_prc: 0.9982
Epoch 16/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0071 - tp: 23005643.9516 - fp: 62651.8063 - tn: 23005643.9516 - fn: 62651.8063 - accuracy: 0.9974 - precision: 0.9974 - recall: 0.9974 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0181 - val_tp: 13822025.0000 - val_fp: 71607.0000 - val_tn: 13822025.0000 - val_fn: 71607.0000 - val_accuracy: 0.9948 - val_precision: 0.9948 - val_recall: 0.9948 - val_auc: 0.9989 - val_prc: 0.9986
Epoch 17/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0066 - tp: 23008332.0627 - fp: 59972.3675 - tn: 23008332.0627 - fn: 59972.3675 - accuracy: 0.9975 - precision: 0.9975 - recall: 0.9975 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0280 - val_tp: 13812753.0000 - val_fp: 80879.0000 - val_tn: 13812753.0000 - val_fn: 80879.0000 - val_accuracy: 0.9942 - val_precision: 0.9942 - val_recall: 0.9942 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 18/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0070 - tp: 23010133.3362 - fp: 58165.4530 - tn: 23010133.3362 - fn: 58165.4530 - accuracy: 0.9973 - precision: 0.9973 - recall: 0.9973 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0162 - val_tp: 13831490.0000 - val_fp: 62142.0000 - val_tn: 13831490.0000 - val_fn: 62142.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9989 - val_prc: 0.9986
Epoch 19/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0069 - tp: 23004443.0741 - fp: 63854.1339 - tn: 23004443.0741 - fn: 63854.1339 - accuracy: 0.9974 - precision: 0.9974 - recall: 0.9974 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0210 - val_tp: 13826265.0000 - val_fp: 67367.0000 - val_tn: 13826265.0000 - val_fn: 67367.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9983 - val_prc: 0.9978
Epoch 20/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0076 - tp: 23000050.3248 - fp: 68246.1425 - tn: 23000050.3248 - fn: 68246.1425 - accuracy: 0.9973 - precision: 0.9973 - recall: 0.9973 - auc: 0.9998 - prc: 0.9998 - val_loss: 0.0169 - val_tp: 13827770.0000 - val_fp: 65862.0000 - val_tn: 13827770.0000 - val_fn: 65862.0000 - val_accuracy: 0.9953 - val_precision: 0.9953 - val_recall: 0.9953 - val_auc: 0.9990 - val_prc: 0.9986
Epoch 21/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0068 - tp: 23011514.1538 - fp: 56780.3476 - tn: 23011514.1538 - fn: 56780.3476 - accuracy: 0.9975 - precision: 0.9975 - recall: 0.9975 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0177 - val_tp: 13828857.0000 - val_fp: 64775.0000 - val_tn: 13828857.0000 - val_fn: 64775.0000 - val_accuracy: 0.9953 - val_precision: 0.9953 - val_recall: 0.9953 - val_auc: 0.9988 - val_prc: 0.9985
Epoch 22/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0045 - tp: 23027395.9288 - fp: 40901.3134 - tn: 23027395.9288 - fn: 40901.3134 - accuracy: 0.9983 - precision: 0.9983 - recall: 0.9983 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0185 - val_tp: 13832646.0000 - val_fp: 60986.0000 - val_tn: 13832646.0000 - val_fn: 60986.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9986 - val_prc: 0.9981
Epoch 23/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0052 - tp: 23024140.1225 - fp: 44158.3989 - tn: 23024140.1225 - fn: 44158.3989 - accuracy: 0.9980 - precision: 0.9980 - recall: 0.9980 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0188 - val_tp: 13829495.0000 - val_fp: 64137.0000 - val_tn: 13829495.0000 - val_fn: 64137.0000 - val_accuracy: 0.9954 - val_precision: 0.9954 - val_recall: 0.9954 - val_auc: 0.9986 - val_prc: 0.9982
Epoch 24/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0047 - tp: 23026346.6410 - fp: 41950.1311 - tn: 23026346.6410 - fn: 41950.1311 - accuracy: 0.9982 - precision: 0.9982 - recall: 0.9982 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0307 - val_tp: 13758166.0000 - val_fp: 135466.0000 - val_tn: 13758166.0000 - val_fn: 135466.0000 - val_accuracy: 0.9902 - val_precision: 0.9902 - val_recall: 0.9902 - val_auc: 0.9985 - val_prc: 0.9981
Epoch 25/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0049 - tp: 23025842.0826 - fp: 42453.7322 - tn: 23025842.0826 - fn: 42453.7322 - accuracy: 0.9982 - precision: 0.9982 - recall: 0.9982 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0192 - val_tp: 13834029.0000 - val_fp: 59603.0000 - val_tn: 13834029.0000 - val_fn: 59603.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9985 - val_prc: 0.9980
Epoch 26/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0043 - tp: 23028640.0883 - fp: 39660.2764 - tn: 23028640.0883 - fn: 39660.2764 - accuracy: 0.9983 - precision: 0.9983 - recall: 0.9983 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0294 - val_tp: 13815143.0000 - val_fp: 78489.0000 - val_tn: 13815143.0000 - val_fn: 78489.0000 - val_accuracy: 0.9944 - val_precision: 0.9944 - val_recall: 0.9944 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 27/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0038 - tp: 23032865.2678 - fp: 35434.3675 - tn: 23032865.2678 - fn: 35434.3675 - accuracy: 0.9985 - precision: 0.9985 - recall: 0.9985 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0223 - val_tp: 13826868.0000 - val_fp: 66764.0000 - val_tn: 13826868.0000 - val_fn: 66764.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9982 - val_prc: 0.9976
Epoch 28/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0041 - tp: 23032204.5385 - fp: 36090.8632 - tn: 23032204.5385 - fn: 36090.8632 - accuracy: 0.9984 - precision: 0.9984 - recall: 0.9984 - auc: 1.0000 - prc: 0.9999 - val_loss: 0.0210 - val_tp: 13830708.0000 - val_fp: 62924.0000 - val_tn: 13830708.0000 - val_fn: 62924.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9983 - val_prc: 0.9978
Epoch 29/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0041 - tp: 23029299.9487 - fp: 38995.3989 - tn: 23029299.9487 - fn: 38995.3989 - accuracy: 0.9984 - precision: 0.9984 - recall: 0.9984 - auc: 1.0000 - prc: 0.9999 - val_loss: 0.0204 - val_tp: 13825606.0000 - val_fp: 68026.0000 - val_tn: 13825606.0000 - val_fn: 68026.0000 - val_accuracy: 0.9951 - val_precision: 0.9951 - val_recall: 0.9951 - val_auc: 0.9985 - val_prc: 0.9981
Epoch 30/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0043 - tp: 23030295.4729 - fp: 38001.1880 - tn: 23030295.4729 - fn: 38001.1880 - accuracy: 0.9983 - precision: 0.9983 - recall: 0.9983 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0197 - val_tp: 13834921.0000 - val_fp: 58711.0000 - val_tn: 13834921.0000 - val_fn: 58711.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9984 - val_prc: 0.9979
Epoch 31/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0037 - tp: 23034444.3191 - fp: 33853.7550 - tn: 23034444.3191 - fn: 33853.7550 - accuracy: 0.9986 - precision: 0.9986 - recall: 0.9986 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0253 - val_tp: 13778864.0000 - val_fp: 114768.0000 - val_tn: 13778864.0000 - val_fn: 114768.0000 - val_accuracy: 0.9917 - val_precision: 0.9917 - val_recall: 0.9917 - val_auc: 0.9989 - val_prc: 0.9986
Epoch 32/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0078 - tp: 23006097.7151 - fp: 62201.4074 - tn: 23006097.7151 - fn: 62201.4074 - accuracy: 0.9969 - precision: 0.9969 - recall: 0.9969 - auc: 0.9999 - prc: 0.9998 - val_loss: 0.0206 - val_tp: 13831133.0000 - val_fp: 62499.0000 - val_tn: 13831133.0000 - val_fn: 62499.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9984 - val_prc: 0.9979
Epoch 33/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0030 - tp: 23040354.5356 - fp: 27942.1709 - tn: 23040354.5356 - fn: 27942.1709 - accuracy: 0.9988 - precision: 0.9988 - recall: 0.9988 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0208 - val_tp: 13834861.0000 - val_fp: 58771.0000 - val_tn: 13834861.0000 - val_fn: 58771.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9983 - val_prc: 0.9978
Epoch 34/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0028 - tp: 23041842.4416 - fp: 26455.0000 - tn: 23041842.4416 - fn: 26455.0000 - accuracy: 0.9989 - precision: 0.9989 - recall: 0.9989 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0199 - val_tp: 13831666.0000 - val_fp: 61966.0000 - val_tn: 13831666.0000 - val_fn: 61966.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9985 - val_prc: 0.9980
Epoch 35/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0024 - tp: 23043482.5926 - fp: 24817.9630 - tn: 23043482.5926 - fn: 24817.9630 - accuracy: 0.9991 - precision: 0.9991 - recall: 0.9991 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0200 - val_tp: 13835736.0000 - val_fp: 57896.0000 - val_tn: 13835736.0000 - val_fn: 57896.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9984 - val_prc: 0.9979
Epoch 36/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0027 - tp: 23042387.4815 - fp: 25909.9060 - tn: 23042387.4815 - fn: 25909.9060 - accuracy: 0.9990 - precision: 0.9990 - recall: 0.9990 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0265 - val_tp: 13827257.0000 - val_fp: 66375.0000 - val_tn: 13827257.0000 - val_fn: 66375.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 37/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0028 - tp: 23041855.1652 - fp: 26438.5755 - tn: 23041855.1652 - fn: 26438.5755 - accuracy: 0.9989 - precision: 0.9989 - recall: 0.9989 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0386 - val_tp: 13815414.0000 - val_fp: 78218.0000 - val_tn: 13815414.0000 - val_fn: 78218.0000 - val_accuracy: 0.9944 - val_precision: 0.9944 - val_recall: 0.9944 - val_auc: 0.9966 - val_prc: 0.9956
Epoch 38/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0034 - tp: 23036199.9630 - fp: 32100.5812 - tn: 23036199.9630 - fn: 32100.5812 - accuracy: 0.9987 - precision: 0.9987 - recall: 0.9987 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0236 - val_tp: 13830426.0000 - val_fp: 63206.0000 - val_tn: 13830426.0000 - val_fn: 63206.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9980 - val_prc: 0.9974
Epoch 39/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0030 - tp: 23038890.1026 - fp: 29407.9031 - tn: 23038890.1026 - fn: 29407.9031 - accuracy: 0.9988 - precision: 0.9988 - recall: 0.9988 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0223 - val_tp: 13827541.0000 - val_fp: 66091.0000 - val_tn: 13827541.0000 - val_fn: 66091.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9983 - val_prc: 0.9978
Epoch 40/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0031 - tp: 23040561.9972 - fp: 27730.7550 - tn: 23040561.9972 - fn: 27730.7550 - accuracy: 0.9988 - precision: 0.9988 - recall: 0.9988 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0256 - val_tp: 13828643.0000 - val_fp: 64989.0000 - val_tn: 13828643.0000 - val_fn: 64989.0000 - val_accuracy: 0.9953 - val_precision: 0.9953 - val_recall: 0.9953 - val_auc: 0.9979 - val_prc: 0.9972
Epoch 41/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0032 - tp: 23039664.1595 - fp: 28635.0199 - tn: 23039664.1595 - fn: 28635.0199 - accuracy: 0.9988 - precision: 0.9988 - recall: 0.9988 - auc: 1.0000 - prc: 0.9999 - val_loss: 0.0250 - val_tp: 13827403.0000 - val_fp: 66229.0000 - val_tn: 13827403.0000 - val_fn: 66229.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9981 - val_prc: 0.9975
Epoch 42/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0029 - tp: 23043745.1425 - fp: 24551.0114 - tn: 23043745.1425 - fn: 24551.0114 - accuracy: 0.9989 - precision: 0.9989 - recall: 0.9989 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0262 - val_tp: 13827186.0000 - val_fp: 66446.0000 - val_tn: 13827186.0000 - val_fn: 66446.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9978 - val_prc: 0.9972
Epoch 43/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0025 - tp: 23045820.5100 - fp: 22478.5499 - tn: 23045820.5100 - fn: 22478.5499 - accuracy: 0.9990 - precision: 0.9990 - recall: 0.9990 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0255 - val_tp: 13832600.0000 - val_fp: 61032.0000 - val_tn: 13832600.0000 - val_fn: 61032.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9979 - val_prc: 0.9972
Epoch 44/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0026 - tp: 23042796.3276 - fp: 25497.3105 - tn: 23042796.3276 - fn: 25497.3105 - accuracy: 0.9989 - precision: 0.9989 - recall: 0.9989 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0238 - val_tp: 13834085.0000 - val_fp: 59547.0000 - val_tn: 13834085.0000 - val_fn: 59547.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9981 - val_prc: 0.9975
Epoch 45/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0027 - tp: 23045045.3362 - fp: 23255.7778 - tn: 23045045.3362 - fn: 23255.7778 - accuracy: 0.9989 - precision: 0.9989 - recall: 0.9989 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0248 - val_tp: 13829666.0000 - val_fp: 63966.0000 - val_tn: 13829666.0000 - val_fn: 63966.0000 - val_accuracy: 0.9954 - val_precision: 0.9954 - val_recall: 0.9954 - val_auc: 0.9981 - val_prc: 0.9975
Epoch 46/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0025 - tp: 23045198.4160 - fp: 23102.0940 - tn: 23045198.4160 - fn: 23102.0940 - accuracy: 0.9990 - precision: 0.9990 - recall: 0.9990 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0250 - val_tp: 13834870.0000 - val_fp: 58762.0000 - val_tn: 13834870.0000 - val_fn: 58762.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9980 - val_prc: 0.9974
Epoch 47/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0025 - tp: 23044539.4986 - fp: 23763.3761 - tn: 23044539.4986 - fn: 23763.3761 - accuracy: 0.9990 - precision: 0.9990 - recall: 0.9990 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0236 - val_tp: 13832076.0000 - val_fp: 61556.0000 - val_tn: 13832076.0000 - val_fn: 61556.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9981 - val_prc: 0.9976
Epoch 48/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0024 - tp: 23045836.9145 - fp: 22459.5214 - tn: 23045836.9145 - fn: 22459.5214 - accuracy: 0.9990 - precision: 0.9990 - recall: 0.9990 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0315 - val_tp: 13828188.0000 - val_fp: 65444.0000 - val_tn: 13828188.0000 - val_fn: 65444.0000 - val_accuracy: 0.9953 - val_precision: 0.9953 - val_recall: 0.9953 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 49/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0025 - tp: 23046585.0285 - fp: 21713.2678 - tn: 23046585.0285 - fn: 21713.2678 - accuracy: 0.9990 - precision: 0.9990 - recall: 0.9990 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0274 - val_tp: 13829900.0000 - val_fp: 63732.0000 - val_tn: 13829900.0000 - val_fn: 63732.0000 - val_accuracy: 0.9954 - val_precision: 0.9954 - val_recall: 0.9954 - val_auc: 0.9977 - val_prc: 0.9969
Epoch 50/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0021 - tp: 23050115.8519 - fp: 18188.2593 - tn: 23050115.8519 - fn: 18188.2593 - accuracy: 0.9992 - precision: 0.9992 - recall: 0.9992 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0256 - val_tp: 13834348.0000 - val_fp: 59284.0000 - val_tn: 13834348.0000 - val_fn: 59284.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9979 - val_prc: 0.9973
Epoch 51/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0019 - tp: 23049435.1994 - fp: 18863.9430 - tn: 23049435.1994 - fn: 18863.9430 - accuracy: 0.9992 - precision: 0.9992 - recall: 0.9992 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0272 - val_tp: 13833493.0000 - val_fp: 60139.0000 - val_tn: 13833493.0000 - val_fn: 60139.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9978 - val_prc: 0.9971
Epoch 52/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0018 - tp: 23050533.6980 - fp: 17763.8746 - tn: 23050533.6980 - fn: 17763.8746 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0355 - val_tp: 13821694.0000 - val_fp: 71938.0000 - val_tn: 13821694.0000 - val_fn: 71938.0000 - val_accuracy: 0.9948 - val_precision: 0.9948 - val_recall: 0.9948 - val_auc: 0.9972 - val_prc: 0.9963
Epoch 53/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0051 - tp: 23035998.8775 - fp: 32301.2821 - tn: 23035998.8775 - fn: 32301.2821 - accuracy: 0.9983 - precision: 0.9983 - recall: 0.9983 - auc: 0.9998 - prc: 0.9998 - val_loss: 0.0260 - val_tp: 13834753.0000 - val_fp: 58879.0000 - val_tn: 13834753.0000 - val_fn: 58879.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9978 - val_prc: 0.9972
Epoch 54/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0021 - tp: 23049337.4017 - fp: 18957.5726 - tn: 23049337.4017 - fn: 18957.5726 - accuracy: 0.9992 - precision: 0.9992 - recall: 0.9992 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0280 - val_tp: 13831323.0000 - val_fp: 62309.0000 - val_tn: 13831323.0000 - val_fn: 62309.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 55/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0018 - tp: 23051160.9972 - fp: 17134.1140 - tn: 23051160.9972 - fn: 17134.1140 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0242 - val_tp: 13833833.0000 - val_fp: 59799.0000 - val_tn: 13833833.0000 - val_fn: 59799.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9981 - val_prc: 0.9975
Epoch 56/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0027 - tp: 23047081.8462 - fp: 21219.7607 - tn: 23047081.8462 - fn: 21219.7607 - accuracy: 0.9990 - precision: 0.9990 - recall: 0.9990 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0287 - val_tp: 13832033.0000 - val_fp: 61599.0000 - val_tn: 13832033.0000 - val_fn: 61599.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 57/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0020 - tp: 23049286.7293 - fp: 19006.7151 - tn: 23049286.7293 - fn: 19006.7151 - accuracy: 0.9992 - precision: 0.9992 - recall: 0.9992 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0238 - val_tp: 13834849.0000 - val_fp: 58783.0000 - val_tn: 13834849.0000 - val_fn: 58783.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9982 - val_prc: 0.9977
Epoch 58/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0024 - tp: 23046874.4558 - fp: 21426.5698 - tn: 23046874.4558 - fn: 21426.5698 - accuracy: 0.9991 - precision: 0.9991 - recall: 0.9991 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0267 - val_tp: 13832569.0000 - val_fp: 61063.0000 - val_tn: 13832569.0000 - val_fn: 61063.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9978 - val_prc: 0.9972
Epoch 59/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0028 - tp: 23041786.3903 - fp: 26514.6211 - tn: 23041786.3903 - fn: 26514.6211 - accuracy: 0.9989 - precision: 0.9989 - recall: 0.9989 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0266 - val_tp: 13829334.0000 - val_fp: 64298.0000 - val_tn: 13829334.0000 - val_fn: 64298.0000 - val_accuracy: 0.9954 - val_precision: 0.9954 - val_recall: 0.9954 - val_auc: 0.9978 - val_prc: 0.9972
Epoch 60/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0016 - tp: 23053253.5442 - fp: 15049.1538 - tn: 23053253.5442 - fn: 15049.1538 - accuracy: 0.9994 - precision: 0.9994 - recall: 0.9994 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0275 - val_tp: 13829799.0000 - val_fp: 63833.0000 - val_tn: 13829799.0000 - val_fn: 63833.0000 - val_accuracy: 0.9954 - val_precision: 0.9954 - val_recall: 0.9954 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 61/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0014 - tp: 23054884.8433 - fp: 13411.0427 - tn: 23054884.8433 - fn: 13411.0427 - accuracy: 0.9994 - precision: 0.9994 - recall: 0.9994 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0316 - val_tp: 13830584.0000 - val_fp: 63048.0000 - val_tn: 13830584.0000 - val_fn: 63048.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9972 - val_prc: 0.9964
Epoch 62/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0017 - tp: 23052942.6040 - fp: 15353.0114 - tn: 23052942.6040 - fn: 15353.0114 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0311 - val_tp: 13826997.0000 - val_fp: 66635.0000 - val_tn: 13826997.0000 - val_fn: 66635.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 63/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0018 - tp: 23051397.5271 - fp: 16902.2222 - tn: 23051397.5271 - fn: 16902.2222 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0285 - val_tp: 13835295.0000 - val_fp: 58337.0000 - val_tn: 13835295.0000 - val_fn: 58337.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 64/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0017 - tp: 23052852.1538 - fp: 15445.8262 - tn: 23052852.1538 - fn: 15445.8262 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0287 - val_tp: 13833756.0000 - val_fp: 59876.0000 - val_tn: 13833756.0000 - val_fn: 59876.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 65/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0017 - tp: 23052019.0969 - fp: 16276.1624 - tn: 23052019.0969 - fn: 16276.1624 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0294 - val_tp: 13833170.0000 - val_fp: 60462.0000 - val_tn: 13833170.0000 - val_fn: 60462.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 66/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0017 - tp: 23052907.9601 - fp: 15392.5726 - tn: 23052907.9601 - fn: 15392.5726 - accuracy: 0.9994 - precision: 0.9994 - recall: 0.9994 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0294 - val_tp: 13832292.0000 - val_fp: 61340.0000 - val_tn: 13832292.0000 - val_fn: 61340.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 67/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0018 - tp: 23050719.7066 - fp: 17585.8974 - tn: 23050719.7066 - fn: 17585.8974 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0276 - val_tp: 13833422.0000 - val_fp: 60210.0000 - val_tn: 13833422.0000 - val_fn: 60210.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9979 - val_prc: 0.9972
Epoch 68/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0016 - tp: 23052423.3960 - fp: 15877.8946 - tn: 23052423.3960 - fn: 15877.8946 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0268 - val_tp: 13833958.0000 - val_fp: 59674.0000 - val_tn: 13833958.0000 - val_fn: 59674.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9979 - val_prc: 0.9972
Epoch 69/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0017 - tp: 23052488.8946 - fp: 15810.9687 - tn: 23052488.8946 - fn: 15810.9687 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0295 - val_tp: 13834854.0000 - val_fp: 58778.0000 - val_tn: 13834854.0000 - val_fn: 58778.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 70/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0016 - tp: 23054135.4615 - fp: 14166.4160 - tn: 23054135.4615 - fn: 14166.4160 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0315 - val_tp: 13832914.0000 - val_fp: 60718.0000 - val_tn: 13832914.0000 - val_fn: 60718.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9975 - val_prc: 0.9967
Epoch 71/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0020 - tp: 23052050.2308 - fp: 16251.3932 - tn: 23052050.2308 - fn: 16251.3932 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 0.9999 - val_loss: 0.0289 - val_tp: 13833857.0000 - val_fp: 59775.0000 - val_tn: 13833857.0000 - val_fn: 59775.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9978 - val_prc: 0.9971
Epoch 72/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0014 - tp: 23054465.3276 - fp: 13835.5812 - tn: 23054465.3276 - fn: 13835.5812 - accuracy: 0.9994 - precision: 0.9994 - recall: 0.9994 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0340 - val_tp: 13830952.0000 - val_fp: 62680.0000 - val_tn: 13830952.0000 - val_fn: 62680.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9973 - val_prc: 0.9964
Epoch 73/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0018 - tp: 23050945.4843 - fp: 17354.6724 - tn: 23050945.4843 - fn: 17354.6724 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0286 - val_tp: 13834117.0000 - val_fp: 59515.0000 - val_tn: 13834117.0000 - val_fn: 59515.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 74/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0022 - tp: 23049529.8519 - fp: 18772.3704 - tn: 23049529.8519 - fn: 18772.3704 - accuracy: 0.9991 - precision: 0.9991 - recall: 0.9991 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0286 - val_tp: 13834577.0000 - val_fp: 59055.0000 - val_tn: 13834577.0000 - val_fn: 59055.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9977 - val_prc: 0.9969
Epoch 75/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0014 - tp: 23056279.5670 - fp: 12023.8063 - tn: 23056279.5670 - fn: 12023.8063 - accuracy: 0.9995 - precision: 0.9995 - recall: 0.9995 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0328 - val_tp: 13830616.0000 - val_fp: 63016.0000 - val_tn: 13830616.0000 - val_fn: 63016.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 76/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0016 - tp: 23053125.7123 - fp: 15177.6296 - tn: 23053125.7123 - fn: 15177.6296 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0302 - val_tp: 13835463.0000 - val_fp: 58169.0000 - val_tn: 13835463.0000 - val_fn: 58169.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 77/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0015 - tp: 23053751.6866 - fp: 14550.5926 - tn: 23053751.6866 - fn: 14550.5926 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0357 - val_tp: 13831430.0000 - val_fp: 62202.0000 - val_tn: 13831430.0000 - val_fn: 62202.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9971 - val_prc: 0.9963
Epoch 78/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0011 - tp: 23057382.5185 - fp: 10913.2222 - tn: 23057382.5185 - fn: 10913.2222 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0301 - val_tp: 13837190.0000 - val_fp: 56442.0000 - val_tn: 13837190.0000 - val_fn: 56442.0000 - val_accuracy: 0.9959 - val_precision: 0.9959 - val_recall: 0.9959 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 79/100
350/350 [==============================] - 19s 54ms/step - loss: 9.0958e-04 - tp: 23058350.0456 - fp: 9948.0655 - tn: 23058350.0456 - fn: 9948.0655 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0360 - val_tp: 13833045.0000 - val_fp: 60587.0000 - val_tn: 13833045.0000 - val_fn: 60587.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 80/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0021 - tp: 23048749.5499 - fp: 19548.5328 - tn: 23048749.5499 - fn: 19548.5328 - accuracy: 0.9992 - precision: 0.9992 - recall: 0.9992 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0266 - val_tp: 13832122.0000 - val_fp: 61510.0000 - val_tn: 13832122.0000 - val_fn: 61510.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9979 - val_prc: 0.9973
Epoch 81/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0013 - tp: 23054806.8490 - fp: 13492.2108 - tn: 23054806.8490 - fn: 13492.2108 - accuracy: 0.9995 - precision: 0.9995 - recall: 0.9995 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0502 - val_tp: 13801970.0000 - val_fp: 91662.0000 - val_tn: 13801970.0000 - val_fn: 91662.0000 - val_accuracy: 0.9934 - val_precision: 0.9934 - val_recall: 0.9934 - val_auc: 0.9961 - val_prc: 0.9950
Epoch 82/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0043 - tp: 23038133.9202 - fp: 30163.7550 - tn: 23038133.9202 - fn: 30163.7550 - accuracy: 0.9983 - precision: 0.9983 - recall: 0.9983 - auc: 0.9999 - prc: 0.9999 - val_loss: 0.0288 - val_tp: 13832938.0000 - val_fp: 60694.0000 - val_tn: 13832938.0000 - val_fn: 60694.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9977 - val_prc: 0.9970
Epoch 83/100
350/350 [==============================] - 19s 54ms/step - loss: 9.3866e-04 - tp: 23059569.5983 - fp: 8732.3248 - tn: 23059569.5983 - fn: 8732.3248 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0322 - val_tp: 13833636.0000 - val_fp: 59996.0000 - val_tn: 13833636.0000 - val_fn: 59996.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9974 - val_prc: 0.9966
Epoch 84/100
350/350 [==============================] - 19s 54ms/step - loss: 8.3007e-04 - tp: 23060483.7265 - fp: 7817.0427 - tn: 23060483.7265 - fn: 7817.0427 - accuracy: 0.9997 - precision: 0.9997 - recall: 0.9997 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0339 - val_tp: 13834353.0000 - val_fp: 59279.0000 - val_tn: 13834353.0000 - val_fn: 59279.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 85/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0011 - tp: 23057801.6154 - fp: 10494.5670 - tn: 23057801.6154 - fn: 10494.5670 - accuracy: 0.9995 - precision: 0.9995 - recall: 0.9995 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0361 - val_tp: 13833064.0000 - val_fp: 60568.0000 - val_tn: 13833064.0000 - val_fn: 60568.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9971 - val_prc: 0.9962
Epoch 86/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0018 - tp: 23049871.0684 - fp: 18429.9915 - tn: 23049871.0684 - fn: 18429.9915 - accuracy: 0.9993 - precision: 0.9993 - recall: 0.9993 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0272 - val_tp: 13832807.0000 - val_fp: 60825.0000 - val_tn: 13832807.0000 - val_fn: 60825.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9979 - val_prc: 0.9972
Epoch 87/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0015 - tp: 23053624.8148 - fp: 14673.9544 - tn: 23053624.8148 - fn: 14673.9544 - accuracy: 0.9994 - precision: 0.9994 - recall: 0.9994 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0292 - val_tp: 13833670.0000 - val_fp: 59962.0000 - val_tn: 13833670.0000 - val_fn: 59962.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 88/100
350/350 [==============================] - 19s 54ms/step - loss: 8.6202e-04 - tp: 23060496.1168 - fp: 7802.0541 - tn: 23060496.1168 - fn: 7802.0541 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0330 - val_tp: 13834355.0000 - val_fp: 59277.0000 - val_tn: 13834355.0000 - val_fn: 59277.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 89/100
350/350 [==============================] - 19s 54ms/step - loss: 7.9374e-04 - tp: 23060107.1368 - fp: 8191.0570 - tn: 23060107.1368 - fn: 8191.0570 - accuracy: 0.9997 - precision: 0.9997 - recall: 0.9997 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0400 - val_tp: 13828450.0000 - val_fp: 65182.0000 - val_tn: 13828450.0000 - val_fn: 65182.0000 - val_accuracy: 0.9953 - val_precision: 0.9953 - val_recall: 0.9953 - val_auc: 0.9968 - val_prc: 0.9958
Epoch 90/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0011 - tp: 23057818.3020 - fp: 10478.2650 - tn: 23057818.3020 - fn: 10478.2650 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0390 - val_tp: 13830219.0000 - val_fp: 63413.0000 - val_tn: 13830219.0000 - val_fn: 63413.0000 - val_accuracy: 0.9954 - val_precision: 0.9954 - val_recall: 0.9954 - val_auc: 0.9969 - val_prc: 0.9959
Epoch 91/100
350/350 [==============================] - 19s 54ms/step - loss: 9.5929e-04 - tp: 23058513.6752 - fp: 9785.1567 - tn: 23058513.6752 - fn: 9785.1567 - accuracy: 0.9995 - precision: 0.9995 - recall: 0.9995 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0299 - val_tp: 13837662.0000 - val_fp: 55970.0000 - val_tn: 13837662.0000 - val_fn: 55970.0000 - val_accuracy: 0.9960 - val_precision: 0.9960 - val_recall: 0.9960 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 92/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0012 - tp: 23057738.7208 - fp: 10557.9772 - tn: 23057738.7208 - fn: 10557.9772 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0323 - val_tp: 13835102.0000 - val_fp: 58530.0000 - val_tn: 13835102.0000 - val_fn: 58530.0000 - val_accuracy: 0.9958 - val_precision: 0.9958 - val_recall: 0.9958 - val_auc: 0.9975 - val_prc: 0.9967
Epoch 93/100
350/350 [==============================] - 19s 54ms/step - loss: 9.0946e-04 - tp: 23059189.1425 - fp: 9106.7379 - tn: 23059189.1425 - fn: 9106.7379 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0298 - val_tp: 13838616.0000 - val_fp: 55016.0000 - val_tn: 13838616.0000 - val_fn: 55016.0000 - val_accuracy: 0.9960 - val_precision: 0.9960 - val_recall: 0.9960 - val_auc: 0.9976 - val_prc: 0.9969
Epoch 94/100
350/350 [==============================] - 19s 54ms/step - loss: 9.7204e-04 - tp: 23058642.2137 - fp: 9657.9088 - tn: 23058642.2137 - fn: 9657.9088 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0345 - val_tp: 13833457.0000 - val_fp: 60175.0000 - val_tn: 13833457.0000 - val_fn: 60175.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9973 - val_prc: 0.9965
Epoch 95/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0013 - tp: 23056930.7550 - fp: 11368.8433 - tn: 23056930.7550 - fn: 11368.8433 - accuracy: 0.9995 - precision: 0.9995 - recall: 0.9995 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0378 - val_tp: 13831474.0000 - val_fp: 62158.0000 - val_tn: 13831474.0000 - val_fn: 62158.0000 - val_accuracy: 0.9955 - val_precision: 0.9955 - val_recall: 0.9955 - val_auc: 0.9971 - val_prc: 0.9962
Epoch 96/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0012 - tp: 23056596.0256 - fp: 11701.0484 - tn: 23056596.0256 - fn: 11701.0484 - accuracy: 0.9995 - precision: 0.9995 - recall: 0.9995 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0349 - val_tp: 13834550.0000 - val_fp: 59082.0000 - val_tn: 13834550.0000 - val_fn: 59082.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9974 - val_prc: 0.9965
Epoch 97/100
350/350 [==============================] - 19s 54ms/step - loss: 8.9556e-04 - tp: 23059321.4786 - fp: 8979.7322 - tn: 23059321.4786 - fn: 8979.7322 - accuracy: 0.9996 - precision: 0.9996 - recall: 0.9996 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0365 - val_tp: 13833508.0000 - val_fp: 60124.0000 - val_tn: 13833508.0000 - val_fn: 60124.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9973 - val_prc: 0.9964
Epoch 98/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0018 - tp: 23048939.5413 - fp: 19360.7322 - tn: 23048939.5413 - fn: 19360.7322 - accuracy: 0.9994 - precision: 0.9994 - recall: 0.9994 - auc: 1.0000 - prc: 0.9999 - val_loss: 0.0349 - val_tp: 13826266.0000 - val_fp: 67366.0000 - val_tn: 13826266.0000 - val_fn: 67366.0000 - val_accuracy: 0.9952 - val_precision: 0.9952 - val_recall: 0.9952 - val_auc: 0.9972 - val_prc: 0.9964
Epoch 99/100
350/350 [==============================] - 19s 54ms/step - loss: 0.0013 - tp: 23056837.3533 - fp: 11460.7407 - tn: 23056837.3533 - fn: 11460.7407 - accuracy: 0.9995 - precision: 0.9995 - recall: 0.9995 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0327 - val_tp: 13832649.0000 - val_fp: 60983.0000 - val_tn: 13832649.0000 - val_fn: 60983.0000 - val_accuracy: 0.9956 - val_precision: 0.9956 - val_recall: 0.9956 - val_auc: 0.9974 - val_prc: 0.9966
Epoch 100/100
350/350 [==============================] - 19s 54ms/step - loss: 6.7551e-04 - tp: 23061854.0826 - fp: 6444.1994 - tn: 23061854.0826 - fn: 6444.1994 - accuracy: 0.9997 - precision: 0.9997 - recall: 0.9997 - auc: 1.0000 - prc: 1.0000 - val_loss: 0.0338 - val_tp: 13834296.0000 - val_fp: 59336.0000 - val_tn: 13834296.0000 - val_fn: 59336.0000 - val_accuracy: 0.9957 - val_precision: 0.9957 - val_recall: 0.9957 - val_auc: 0.9974 - val_prc: 0.9965
In [17]:
# Save model for future use
model.save('3D_UNet_patch_normalized.h5')
In [18]:
def plot_metrics(history):
    metrics = ['loss', 'prc', 'precision', 'recall']
    fig = plt.figure(figsize = (8, 8))
    for n, metric in enumerate(metrics):
        name = metric.replace("_"," ").capitalize()
        plt.subplot(2,2,n+1)
        plt.plot(history.epoch, history.history[metric], color='blue', label='Train')
        plt.plot(history.epoch, history.history['val_'+metric],
                 color='blue', linestyle="--", label='Val')
        plt.xlabel('Epoch')
        plt.ylabel(name)
        if metric == 'loss':
            plt.ylim([0, plt.ylim()[1]])
        elif metric == 'auc':
            plt.ylim([0.8,1])
        else:
            plt.ylim([0,1])

        plt.legend();
In [19]:
plot_metrics(history)
In [6]:
#Load the pretrained model for testing and predictions. 
from tensorflow.keras.models import load_model
my_model = load_model('3D_UNet_patch_normalized.h5', compile=False)
In [7]:
''' Predict on Training Set '''
img = x_train_orig.copy()
ground_truth = y_train_orig.copy()

# convert training images to patches
img, ground_truth = to_patches(img, ground_truth)
# predict
train_predictions_baseline = my_model.predict(img)
# argmax
train_prediction = np.argmax(train_predictions_baseline, axis=4)[:,:,:,:]

''' Stitch the patches back into Images '''
end = []
gt = []
for x in range(67):
    temp_img = train_prediction[(64*x):64*(x+1)]
    temp_img = patches_to_img(temp_img)
    end.append(temp_img)
for x in range(67):
    temp_img = ground_truth[(64*x):64*(x+1)]
    temp_img = patches_to_img(temp_img)
    gt.append(temp_img)
    
train_predictions = np.asarray(end)
print(train_predictions.shape)

(67, 256, 256, 128)
In [8]:
''' Mean IoU on Training Set '''
from tensorflow.keras.metrics import MeanIoU

n_classes = 2
IOU_keras = MeanIoU(num_classes=n_classes) 

gt1 = gt
IOU_keras.update_state(gt1, train_predictions)
print("Training Set:", IOU_keras.result().numpy())

Training Set: 0.81324434
In [9]:
''' Training Set Per Pixel Basis '''
from sklearn.metrics import confusion_matrix
import seaborn as sns
y_train_matrix = confusion_matrix(np.asarray(gt1).flatten(), np.asarray(train_predictions).flatten())

ax= plt.subplot()
sns.heatmap(y_train_matrix, annot=True, fmt='d', ax=ax);  #annot=True to annotate cells, ftm='g' to disable scientific notation

# labels, title and ticks
ax.set_xlabel('Predicted labels');
ax.set_ylabel('True labels'); 
ax.set_title('Train Set Per Pixel Basis'); 
ax.xaxis.set_ticklabels(['0', '1']); ax.yaxis.set_ticklabels(['0', '1']);
In [10]:
''' Predict on Testing Set '''
img = x_test_orig.copy()
ground_truth = y_test_orig.copy()

# split images to patches
img, ground_truth = to_patches(img, ground_truth)
# predict
test_predictions_baseline = my_model.predict(img)
# argmax
test_prediction = np.argmax(test_predictions_baseline, axis=4)[:,:,:,:]

''' Stitch patches back into Images '''
end = []
gt = []
for x in range(17):
    temp_img = test_prediction[(64*x):64*(x+1)]
    temp_img = patches_to_img(temp_img)
    end.append(temp_img)
for x in range(17):
    temp_img = ground_truth[(64*x):64*(x+1)]
    temp_img = patches_to_img(temp_img)
    gt.append(temp_img)
    
test_predictions = np.asarray(end)
print(test_predictions.shape)

(17, 256, 256, 128)
In [11]:
''' Mean IoU for Testing Set '''
from tensorflow.keras.metrics import MeanIoU

n_classes = 2
IOU_keras = MeanIoU(num_classes=n_classes)  
gt2 = gt
IOU_keras.update_state(gt2, test_predictions)
print("Testing Set:", IOU_keras.result().numpy())

Testing Set: 0.7774364
In [12]:
''' Test Set Per Pixel Basis '''
y_test_matrix = confusion_matrix(np.asarray(gt2).flatten(), np.asarray(test_predictions).flatten())

ax= plt.subplot()
sns.heatmap(y_test_matrix, annot=True, fmt='d', ax=ax);  #annot=True to annotate cells, ftm='g' to disable scientific notation

# labels, title and ticks
ax.set_xlabel('Predicted labels');
ax.set_ylabel('True labels'); 
ax.set_title('Test Set Per Pixel Basis'); 
ax.xaxis.set_ticklabels(['0', '1']); ax.yaxis.set_ticklabels(['0', '1']);
In [ ]:
In [13]:
def plot_three(image, mask, predicted):
    index = 0
    count = 1
    fig = plt.figure(figsize = (12, 500))
    for x in range(1, 129):
        for y in range(1, 4):
            plt.subplot(128, 3, count).axis("off")
            if count % 3 == 1:
                plt.title("Image Slice: " + str(index))
                plt.imshow(image[:,:,index], cmap='gray')
            elif count % 3 == 2:
                plt.title("Mask Slice: " + str(index))
                plt.imshow(mask[:,:,index])
            else:
                plt.title("Predicted Mask Slice: " + str(index))
                plt.imshow(predicted[:,:,index])
                index += 1
            count += 1
In [14]:
plot_three(x_test_orig[3], y_test_orig[3], test_predictions[3])
In [ ]: